In this 51-patient SEEG sample, recalled-word beta power matched the theta-to-low-gamma line; the alpha remainder was inconclusive

Question: the 'information via desynchronization' idea predicts a band-specific alpha/beta power decrease for words that are recalled later. A broadband (aperiodic) change, which 177a70c3 found, also changes alpha and beta band power. Does an alpha/beta decrease stay after we remove a straight line in log-power / log-frequency space? Data: the 54 archived sample-timed feature files of d3aeb47e (OpenNeuro ds004789, FR1 session 0, 0-1.6 s bipolar log band power). 51 subjects had >= 15 recalled and >= 15 not-recalled words. Method (PLAN.md hashed before any statistic): per pair, list and serial-position means removed by OLS; line residual = band power minus the line through theta (3-8 Hz) and low gamma (30-60 Hz) at the band centre; recalled minus not-recalled in dB, mean over pairs; sign-flip test over subjects, Holm over 3. Results: raw beta (13-30 Hz) was lower for recalled words (-0.040 dB, 34/51 subjects, Holm p 0.040). The beta line residual was -0.006 dB (CI -0.030 to +0.019, Holm p 0.66), smaller than raw (paired p 0.018). 3 other beta residuals (3-point line, low gamma/HFA extrapolation, per-pair regression) were also not significant. Raw alpha was lower (-0.091 dB, p 0.0003); its line residual (-0.035 dB, Holm p 0.135) was not significant, but the per-pair regression residual was (-0.053 dB, Holm p 0.014). Thus the beta effect is shared with the neighbouring bands, not beta-specific; a small alpha-specific part is possible. Limits: 5 band means only, no knee; the theta anchor has its own memory effect; one window; effects are small. The package has the plan, numpy code, input SHA-256 values, results and a rerun script that gave identical results.

subsequent memory effect · SEEG · beta power · alpha power · spectral tilt · free recall · broadband spectrum